RAFAEL DAVID JOSÉ

DESCRIPTION

"Trash Picker" is a demonstration game designed to showcase the implementation of a Naive Bayes classifier for creating an AI that learns from human behavior. In this game, you control Luso, a robot tasked with picking up trash in a grid. The game can be played by a human or by an AI that has learned from observing human players.

SCREENSHOTS

MY ROLE & CONTRIBUTION

I implemented the AI system and project documentation. My responsibilities included integrating the Naive Bayes classifier library for machine learning functionality, developing the UI system with buttons for human and AI player modes, and creating the complete project documentation including methodology, results, and analysis. I also handled the leaderboard functionality, player movement implementation, and grid instantiation systems. Additionally, I was responsible for adding comprehensive XML code documentation and creating detailed README documentation covering the AI methodology, UI implementation, and research findings.

CORE CONCEPTS

This project demonstrates the use of supervised machine learning in games, specifically through the application of a Naive Bayes classifier. It explores the process of training an AI based on human input, analyzing patterns in player behavior, and using probability-based decision-making to replicate player actions in a controlled environment.

TECHNICAL ASPECTS

Trash Picker AI was developed in Unity and uses C# for gameplay mechanics and AI integration. The project includes a grid-based movement system and a scoring system based on collected trash. The AI uses a Naive Bayes classifier trained on player input data, allowing it to simulate human behavior after observation. All data collection and prediction logic is implemented through custom scripts without external libraries.